From mboxrd@z Thu Jan 1 00:00:00 1970 From: Kent Overstreet Subject: Re: [PATCH v6 11/13] block: Rework bio_pair_split() Date: Fri, 24 Aug 2012 03:37:34 -0700 Message-ID: <20120824103734.GH11977@moria.home.lan> References: <1345655050-28199-1-git-send-email-koverstreet@google.com> <1345655050-28199-12-git-send-email-koverstreet@google.com> <20120822210410.GL19212@google.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: Sender: linux-bcache-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: "Martin K. Petersen" Cc: Tejun Heo , linux-bcache-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, dm-devel-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org, vgoyal-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org, mpatocka-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org, bharrosh-C4P08NqkoRlBDgjK7y7TUQ@public.gmane.org, Jens Axboe , NeilBrown , Lars Ellenberg , Peter Osterlund , Sage Weil List-Id: dm-devel.ids On Thu, Aug 23, 2012 at 10:25:47PM -0400, Martin K. Petersen wrote: > >>>>> "Tejun" == Tejun Heo writes: > > Tejun> I complained about this in the last posting and in the previous > Tejun> patch. Please respond. Martin, are you okay with these > Tejun> integrity changes? > > I missed the first several iterations of all this while I was out on > vacation. I'll have to try to wrap my head around the new approach. > > However, I'm not sure I like the overall approach of the new splitting. > Instead of all this cloning, slicing and dicing of bio_vecs I'd rather > we bit the bullet and had an offset + length for the vector inside each > bio. That way we could keep the bio_vec immutable and make clones more > lightweight since their vecs would always point to the parent. This also > makes it trivial to split I/Os in the stacking drivers and removes evils > in the partial completion code path. It would also allow to sever the > ties between "size of block range operated on" vs. bi_size which we need > for copy offload, discard, etc. Agree 110% - making bio_vecs immutable and keeping the offset in the bio is something I've been talking about for ages, I'd love to see it happen. But that's going to be a much more invasive change so if I'm going to do it (and I am willing to work on it) it's just going to be a bit. This is really a stopgap solution. As far as the integrity splitting, it's similar to what the existing dm code does (main difference is dm already has the bio cloned, my bio_split() doesn't assume anything about the bio being split). Not sure how that affects ownership of the integrity data, honestly that part kind of confuses me.